Noise from FET controlled by comparator

[DLam]

Hi there!

I have an op amp acting as a comparator (which might not be too ideal) comparing ground and an inverted signal from a coil that, when nothing magnetic is moving near it, is pretty close to ground (±1mV or so). I'm using this comparator (Vcc=7.4V from the +ve terminal of a battery, Vee=-7.4V from the 7660S operating in voltage inverting mode) to drive a FET whose drain is connected to a signal, but my Arduino's ADC was showing some significant noise (LOGGER4.txt: the incoming signal was at ground at that time and I expected to read 0's with small deviations). Originally, I thought it was a problem with the signal (which was also coming from the induced EMF of another coil and can be pretty noisy) but even when I connected the source of the FET to ground and the drain to the ADC with a pull-down resistor (shown above), I was still getting noise, albeit not as severe (LOGGER5.txt, again expecting to read 0's), but this time I have no idea how the ADC could be reading anything other than ground. Any ideas? Thanks in advance!

 

[FvM]

The observed circuit behaviour looks expectable. The comparator threshold is zero without a hysteresis, so it will be switching with no input signal.

I have a different problem with the circuit, I don't understand it's purpose. Why would we send the output of a comparator, amplified by another transistor stage to an ADC?

 

[DLam]

The observed circuit behaviour looks expectable. The comparator threshold is zero without a hysteresis, so it will be switching with no input signal.

Hmm... I see what you're saying, but even if it is switching, wouldn't both high and low output levels cause the ADC to see 0V? I guess I don't understand where any positive voltages could be contaminating the signal to the ADC.

I have a different problem with the circuit, I don't understand it's purpose. Why would we send the output of a comparator, amplified by another transistor stage to an ADC?

Sorry for my lack of transparency. Originally, this circuit was supposed to switch between the non-inverted amplified induced EMF signal from the coil if the induced EMF was positive and the inverted signal if the induced EMF was negative, since the ADC can only read positive voltages. Basically, it was supposed to act as an absolute value circuit. I know others exist, but I just wanted to test the concept out.

This isn't the full circuit, but this was the only part causing me problems. In reality, the output from the comparator is connected to an inverter (operating from 0V-7.4V), which is then connected to the gate of another FET which supplies the inverted signal to the ADC. Strangely, when the ADC was reading the inverted signal, it was far less noisy. I guess the main thing I don't understand is the discrepancy in noise.

 

[mrinalmani]

The reason may be coupling of leakage magnetic field from to coil with the traces of the PCB and thereby inducing voltage at the ADC input. Is the coil close to the ADC?
Another reason may be ringing of the rail voltage due to the coil and some adventitious parasitic capacitance. Place a 10 ohm resistor in series with the coil and check again.

 

[DLam]

mrinalmani, thanks for the suggestion, but there didn't seem to be much effect. I was still getting about the same amount of noise (LOGGER6.txt) The coil is admittedly quite close to the board (~2cm from the edge, ~5cm from the op amp), but could it cause that much noise if nothing magnetic is moving near it and there is no appreciable current flowing through it either?

I forgot to mention that I am also switching the noninverting input of the first op amp between the coil between the coil and ground at about 100Hz and measuring only when the input is connected to the coil. In retrospect doesn't seem very good since the comparator doesn't have a hysteresis; could that be the issue?

One last thing: I reversed the connections to the comparator and sent the output signal through an inverter that outputs from 0V to 7.8V and connected the latter's output to the gate of the FET, and the noise seems to have disappeared. I can't really see why though... it's like the negative gate-source voltage is causing the problem... :|

 

[mrinalmani]

There we have.... "inductor switching at 100KHZ!"
1. The switching of inductor is likely causing the comparator to swing. Note that when you switch-off the inductor, the ground pin of amp number 1 is floating and has possibility of acquiring a small voltage.
2. The swinging of comparator in turn swings the gate voltage of the MOSFET between +7 and -7
3. This oscillating gate voltage oscillates the gate to drain parasitic capacitor of the MOSFET (typical values 300pF)
4. Negative voltage across this parasitic capacitor appears across the AD converter pin.
5. The converter is not designed for negative voltages, and there you have problems.

6. Add a shunt diode across the gate-source pin and check again.

 

[FvM]

but even if it is switching, wouldn't both high and low output levels cause the ADC to see 0V?

Unlikely, because the gate-drain capacitance couples a considerable AC voltage amount to the drain when he comparator is switching. I don't know how the full circuit looks like, but it's basically a bad idea to use a power MOSFET like IRFD120 as analog switch. You better think about CMOS switches with low charge injection.

 

[DLam]

Thanks for all of the helpful suggestions! I guess it makes sense that the capacitance is the culprit: the noise also disappears when I stop switching the coil. I wasn't really aware that the effects could be so macroscopic, but I gladly stand corrected. I have yet to try adding a shunt diode, fine-tuning the hysteresis or implementing a CMOS switch; I'll give them a spin and see what happens.